A team of researchers at Arizona State University (ASU) has unveiled a groundbreaking artificial intelligence tool, Ark+, aimed at enhancing how doctors interpret chest X-rays.
With the promise to “help doctors read chest X-rays better and improve health care outcomes,” Ark+ sets a new benchmark for open, accessible, and effective medical AI.
Built with the vision of democratizing medical technology, Ark+ is designed to be “an open, reliable and ultimately useful tool in real-world health care systems,” explained Jianming “Jimmy” Liang, professor at ASU’s College of Health Solutions and lead author of the study published in Nature.
The innovation comes at a time when patients are demanding more value from healthcare while the U.S. continues to lag in global health rankings, currently 49th in life expectancy, according to the World Bank.
Doctors face pressure to get diagnoses right the first time. Ark+ steps in not just as a diagnostic tool, but as a catalyst for more accurate, faster, and fairer healthcare.
Chest X-rays are among the most common imaging tools in medicine. They help doctors detect conditions like pneumonia, tuberculosis, heart enlargement, and broken ribs. However, interpreting them can be subjective. Even experienced professionals may struggle, especially with rare or emerging diseases like COVID-19 or avian flu.
Ark+ addresses these limitations by reducing diagnostic errors and speeding up interpretation. Even more impressive, it outperformed proprietary tools from tech giants like Google and Microsoft.
The Ark+ model was trained on over 700,000 chest X-ray images sourced from public datasets around the world. But its real edge came from using detailed expert physician notes, something often left out in AI training by larger companies.
“You learn more knowledge from experts,” Liang emphasized. “Ark+ is accruing and reusing knowledge.”
This human-in-the-loop approach enhanced the AI’s performance using fully supervised learning. This method contrasts with the self-supervised models employed by most companies.
By preserving and learning from expert labels, Ark+ delivers more accurate results, especially for underrepresented or rare conditions.
Ark+ was developed by a small but determined team, including ASU graduate students DongAo Ma and Jiaxuan Pang, with funding from the National Institutes of Health, the National Science Foundation, and seed support from the Mayo Clinic Arizona.
Recognizing the challenges of competing directly with industry titans, Liang said, “If we compete directly, it’s unlikely that we’re going to win. But with open-source software, we invite collaborations with many other labs.”
This open approach is their strategic advantage, and may just be the “slingshot” needed to disrupt the dominance of proprietary medical AI systems.
Liang notes that Ark+ can potentially be applied to other imaging technologies like CT scans and MRIs, expanding its utility in modern diagnostics. The team is also looking toward commercializing the tool for hospitals, while keeping it open-source for continued research and innovation.
Their ultimate goal? Making AI-powered diagnostics fairer, smarter, and more accessible, especially for underserved regions.
“By making this model fully open, we’re inviting others to join us in making medical AI more fair, accurate and accessible,” Liang stated. “We believe this will help save lives.”
As the U.S. grapples with a healthcare system that often delivers less for more, innovations like Ark+ offer a much-needed course correction. They make advanced diagnostics more reliable and universally accessible.
It’s a rare case of academic innovation taking on the corporate Goliaths, and maybe, just maybe, giving healthcare a much-needed AI-powered upgrade.